Documentation Grafana Alloy Release notes

Release notes for Grafana Alloy

The release notes provide information about deprecations and breaking changes in Grafana Alloy.

For a complete list of changes to Grafana Alloy, with links to pull requests and related issues when available, refer to the Changelog.

v1.11

Breaking changes due to major version upgrade of Prometheus

Prometheus dependency had a major version upgrade from v2.55.1 to v3.4.2.

• The . pattern in regular expressions in PromQL matches newline characters now. With this change a regular expressions like .* matches strings that include \n. This applies to matchers in queries and relabel configs in Prometheus and Loki components.

• The enable_http2 in prometheus.remote_write component’s endpoints has been changed to false by default. Previously, in Prometheus v2 the remote write http client would default to use http2. In order to parallelize multiple remote write queues across multiple sockets its preferable to not default to http2. If you prefer to use http2 for remote write you must now set enable_http2 to true in your prometheus.remote_write endpoints configuration section.

• The experimental CLI flag --feature.prometheus.metric-validation-scheme has been deprecated and has no effect. You can configure the metric validation scheme individually for each prometheus.scrape component.

• Log message format has changed for some of the prometheus.* components as part of the upgrade to Prometheus v3.

• The values of the le label of classic histograms and the quantile label of summaries are now normalized upon ingestion. In previous Alloy versions, that used Prometheus v2, the value of these labels depended on the scrape protocol (protobuf vs text format) in some situations. This led to label values changing based on the scrape protocol. E.g. a metric exposed as my_classic_hist{le="1"} would be ingested as my_classic_hist{le="1"} via the text format, but as my_classic_hist{le="1.0"} via protobuf. This changed the identity of the metric and caused problems when querying the metric. In current Alloy release, which uses Prometheus v3, these label values will always be normalized to a float like representation. I.e. the above example will always result in my_classic_hist{le="1.0"} being ingested into Prometheus, no matter via which protocol. The effect of this change is that alerts, recording rules and dashboards that directly reference label values as whole numbers such as le="1" will stop working.

  The recommended way to deal with this change is to fix references to integer le and quantile label values, but otherwise do nothing and accept that some queries that span the transition time will produce inaccurate or unexpected results.

See the upstream Prometheus v3 migration guide for more details.

Breaking changes in prometheus.scrape

scrape_native_histograms attribute for prometheus.scrape is now set to false, whereas in previous versions of Alloy it would default to true. This means that it is no longer enough to just configure scrape_protocols to start with PrometheusProto to scrape native histograms - scrape_native_histograms has to be enabled. If scrape_native_histograms is enabled, scrape_protocols will automatically be configured correctly for you to include PrometheusProto. If you configure it explicitly, Alloy will validate that PrometheusProto is in the scrape_protocols list.

In previous versions of Alloy configuring scrape_protocols to start with PrometheusProto was enough to start scraping native histograms because scrape_native_histogram defaulted to true:

alloy Copy

prometheus.scrape "scrape" {
  scrape_protocols = ["PrometheusProto"]
}

Now it has to be enabled and scrape_protocols can be omitted:

alloy Copy

prometheus.scrape "scrape" {
  scrape_native_histograms = true
}

Breaking changes in prometheus.exporter.windows

As the windows_exporter continues to be refactored upstream, there are various breaking changes in metrics.

• windows_process_start_time -> windows_process_start_time_seconds_timestamp in the process collector.
• windows_time_clock_frequency_adjustment_ppb_total -> windows_time_clock_frequency_adjustment_ppb in the time collector.
• windows_net_nic_info -> windows_net_nic_address_info in net collector.
• windows_system_boot_time_timestamp_seconds -> windows_system_boot_time_timestamp in system collector.
• windows_os_physical_memory_free_bytes -> windows_memory_physical_free_bytes from os collector -> memory collector.
• windows_os_process_memory_limit_bytes -> windows_memory_process_memory_limit_bytes from os collector -> memory collector.
• windows_os_processes -> windows_system_processes from os collector -> system collector.
• windows_os_processes_limit -> windows_system_process_limit from os collector -> system collector.
• windows_os_time -> windows_time_current_timestamp_seconds from os collector -> time collector.
• windows_os_timezone -> windows_time_timezone from os collector -> time collector.
• windows_os_users from os collector can be reconstructed by aggregating windows_terminal_services_session_info{state="active"} in terminal_services collector.
• windows_os_virtual_memory_bytes -> windows_memory_commit_limit from os collector -> memory collector.
• windows_os_virtual_memory_free_bytes from os collector can be reconstructed by subtracting windows_memory_committed_bytes from windows_memory_commit_limit in memory collector.
• windows_os_visible_memory_bytes -> windows_memory_physical_total_bytes from os collector -> memory collector.

Deprecated collectors have been removed. Configuration of these collectors will be allowed in Alloy for at least one minor version, but will have no effect.

• logon collector removed, use the terminal_services collector instead.
• cs collector removed, use the os, memory, or cpu collectors instead.

Refer to the release notes for the windows_exporter version v0.31.0 for the breaking change details.

v1.9

Breaking change: The prometheus.exporter.oracledb component now embeds a different exporter

The prometheus.exporter.oracledb component now embeds the oracledb_exporter from oracle instead of the deprecated oracledb_exporter from iamseth.

As a result of this change, the following metrics are no longer available by default:

• oracledb_sessions_activity
• oracledb_tablespace_free_bytes

The previously undocumented argument custom_metrics is now expecting a list of paths to custom metrics files.

Breaking change: The enable_context_propagation argument in beyla.ebpf has been replaced with the context_propagation argument.

Set enable_context_propagation to all to get the same behaviour as enable_context_propagation being set to true.

Breaking change: In prometheus.exporter.windows, the service and msmq collectors no longer work with WMI

The msmq block has been removed. The enable_v2_collector, where_clause, and use_api attributes in the service block are also removed.

Prior to Alloy v1.9.0, the service collector exists in 2 different versions. Version 1 used WMI (Windows Management Instrumentation) to query all services and was able to provide additional information. Version 2 is a more efficient solution by directly connecting to the service manager, but is not able to provide additional information like run_as or start configuration.

In Alloy v1.9.0 the Version 1 collector was removed, hence why some arguments and blocks were removed. In Alloy v1.9.2 those arguments and blocks were re-introduced as a no-op in order to make migrations easier for customers.

Due to this change, the metrics produced by service collector are different in v1.9.0 and above. The msmq collector metrics are unchanged.

Example V2 service metrics:

Copy

windows_service_state{display_name="Declared Configuration(DC) service",name="dcsvc",status="continue pending"} 0
windows_service_state{display_name="Declared Configuration(DC) service",name="dcsvc",status="pause pending"} 0
windows_service_state{display_name="Declared Configuration(DC) service",name="dcsvc",status="paused"} 0
windows_service_state{display_name="Declared Configuration(DC) service",name="dcsvc",status="running"} 0
windows_service_state{display_name="Declared Configuration(DC) service",name="dcsvc",status="start pending"} 0
windows_service_state{display_name="Declared Configuration(DC) service",name="dcsvc",status="stop pending"} 0
windows_service_state{display_name="Declared Configuration(DC) service",name="dcsvc",status="stopped"} 1

For more information on V1 and V2 service metrics, see the upstream exporter documentation for version 0.27.3 of the Windows Exporter, which is the version used in Alloy v1.8.3. Alloy v1.9.2 uses version 0.30.7 of the Windows Exporter.

v1.6

Breaking change: The topics argument in the component loki.source.kafka does not use regex by default anymore

A bug in loki.source.kafka caused the component to treat all topics as regular expressions. For example, setting the topic value to “telemetry” would match any topic containing the substring “telemetry”. With the fix introduced in this version, topic values are now treated as exact matches by default. Regular expression matching is still supported by prefixing a topic with “^”, allowing it to match multiple topics.

Breaking change: Change decision precedence in otelcol.processor.tail_sampling when using and_sub_policy and invert_match

Alloy v1.5 upgraded to OpenTelemetry Collector v0.104.0, which included a fix to the tail sampling processor:

Previously if the decision from a policy evaluation was NotSampled or InvertNotSampled it would return a NotSampled decision regardless, effectively downgrading the result. This was breaking the documented behaviour that inverted decisions should take precedence over all others.

The “documented behavior” which the above quote is referring to is in the processor documentation:

Each policy will result in a decision, and the processor will evaluate them to make a final decision:

• When there’s an “inverted not sample” decision, the trace is not sampled;
• When there’s a “sample” decision, the trace is sampled;
• When there’s a “inverted sample” decision and no “not sample” decisions, the trace is sampled;
• In all other cases, the trace is NOT sampled

An “inverted” decision is the one made based on the “invert_match” attribute, such as the one from the string, numeric or boolean tag policy.

However, in OpenTelemetry Collector v0.116.0 this fix was reverted:

Reverts #33671, allowing for composite policies to specify inverted clauses in conjunction with other policies. This is a change bringing the previous state into place, breaking users who rely on what was introduced as part of #33671.

v1.5

Breaking change: Change default value of max_streams in otelcol.processor.deltatocumulative

The default value was changed from 0 to 9223372036854775807 (max int).

Breaking change: Change default value of namespace in otelcol.connector.spanmetrics

The default value was changed from "" to "traces.span.metrics".

Breaking change: The component otelcol.exporter.logging has been removed in favor of otelcol.exporter.debug

Both components are very similar. More information can be found in the announcement issue.

Breaking change: Change default value of revision in import.git

The default value was changed from "HEAD" to "main". Setting the revision to "HEAD", "FETCH_HEAD", "ORIG_HEAD", "MERGE_HEAD" or "CHERRY_PICK_HEAD" is no longer allowed.

v1.4

Breaking change: Some debug metrics for otelcol components have changed

For example, otelcol.exporter.otlp’s exporter_sent_spans_ratio_total metric is now otelcol_exporter_sent_spans_total. You may need to change your dashboard and alert settings to reference the new metrics. Refer to each component’s documentation page for more information.

Breaking change: The convert_sum_to_gauge and convert_gauge_to_sum functions in otelcol.processor.transform change context

The convert_sum_to_gauge and convert_gauge_to_sum functions must now be used in the metric context rather than in the datapoint context. This is due to a change upstream.

Breaking change: Renamed metrics in beyla.ebpf

process.cpu.state is renamed to cpu.mode and beyla_build_info is renamed to beyla_internal_build_info.

v1.3

Breaking change: remotecfg block updated argument name from metadata to attributes

The remotecfg block has an updated argument name from metadata to attributes.

v1.2

Breaking change: remotecfg block updated for Agent rename

The remotecfg block has been updated to use alloy-remote-config over agent-remote-config. This change aligns remotecfg API terminology with Alloy and includes updated endpoints.